A word recognition score, or WRS, is the percentage of single-syllable words you correctly repeat back during a hearing test. Your audiologist reads or plays a list of short words (like “cat,” “ship,” or “door”) through headphones, and you say each one back as best you can. If you get 42 out of 50 right, your score is 84%. The number sounds simple, but it tells clinicians something a standard hearing test cannot: how well your brain and ears work together to turn sound into meaning. A pure-tone audiogram shows the softest sounds you can detect at different pitches, while the WRS reveals whether turning speech up louder actually helps you understand it.
How the Test Works
The standard procedure uses lists of monosyllabic words, most commonly drawn from two classic sets developed decades ago: the CID W-22 lists and the Northwestern University Auditory Test No. 6 (NU-6). Both contain phonetically balanced words, meaning the speech sounds in each list roughly mirror how often those sounds show up in everyday English. The two tests are not interchangeable, though. Research comparing them found that scores on the W-22 lists tend to come out higher than NU-6 scores at the same loudness levels, so knowing which list was used matters when comparing results over time or between clinics.1PubMed. A comparison of the NU-6 and W-22 speech discrimination tests for assessing sensorineural hearing loss
You typically hear the words through insert earphones in a soundproof booth. Most clinics present either 25 or 50 words per ear. The audiologist scores each word as right or wrong, though some clinics score individual speech sounds (phonemes) within each word, giving a slightly more granular picture. Words are usually presented at a fixed loudness level chosen to give you the best chance of understanding, which brings us to one of the trickiest parts of the test: picking the right volume.
Why the Volume Setting Matters So Much
Audiologists want to find your best possible score, sometimes called “PB Max” (short for phonetically balanced maximum). The idea is to present words loud enough that you can hear all the speech sounds clearly, but not so loud that the signal distorts or becomes uncomfortable. Finding that sweet spot is harder than it sounds.
Research has shown that the range of volumes yielding your best score spans roughly 33 decibels in people with normal hearing, but that window narrows as hearing loss worsens.2PubMed. Range of intensities yielding PB Max and the threshold for monosyllabic words for hearing-impaired subjects For someone with moderate or severe loss, testing at a level just below the loudness discomfort level tends to produce a more accurate estimate of PB Max than testing at a comfortable listening level. That same study recommended getting scores at a minimum of two different volumes so the audiologist can be confident the result reflects your actual ability rather than a level that happened to be too soft or too loud.
This matters in practice because presentation level choices vary across clinics. A study comparing several common methods found that people with moderate or worse hearing loss scored significantly lower when words were presented at their most comfortable level compared to a level calculated from their uncomfortable loudness threshold.3PubMed Central. A Comparison of Presentation Levels to Maximize Word Recognition Scores For mild losses, the presentation method barely mattered. The takeaway: if you have a noticeable hearing loss and your audiologist only tested at one comfortable level, your score may underestimate what you could actually achieve.
Interpreting the Numbers
WRS results are usually grouped into descriptive categories. A score of 90–100% is considered normal or excellent word recognition. Scores in the 76–88% range suggest mild difficulty, 60–74% moderate difficulty, 40–58% poor recognition, and below 40% very poor. These cutoffs are informal guidelines rather than hard diagnostic boundaries, but they give audiologists a quick sense of how much speech understanding you retain.
Raw scores alone, though, only tell part of the story. A more refined approach compares your actual score to a predicted score based on the audibility of speech sounds given your audiogram. This predicted-versus-observed comparison, formalized as a standardized word recognition score (sWRS), expresses how far your performance deviates from what would be expected for someone with your degree of hearing loss.4Ear and Hearing. Clinical Utility of the Standardized Word Recognition Score When the observed score falls well below the predicted one, it raises a red flag: something beyond the audiogram is dragging down your speech understanding.
In a study of older adults, about 12% had at least one word recognition score that fell significantly below their predicted score, which the researchers considered a subtle sign of possible auditory neuropathy, a condition where sound reaches the inner ear fine but the neural signal traveling to the brain gets scrambled.5PubMed. Word recognition and the articulation index in older listeners with probable age-related auditory neuropathy Without comparing observed and predicted scores, that kind of neural deficit can hide behind an otherwise unremarkable audiogram.
Statistical Reliability and the “Critical Difference”
One underappreciated aspect of WRS testing is that scores bounce around a bit from one session to the next, just by chance. If you score 72% today and 80% next month, that might not mean your hearing changed at all. Because each word is scored as right or wrong, WRS data follow a binomial distribution, and the variability is largest in the middle of the scoring range (around 50%) and smallest near the extremes (near 0% or 100%).
Researchers have built “critical difference” tables that tell audiologists how big a change between two scores needs to be before it is statistically meaningful. These tables depend on how many words were presented. With a 25-word list, a change of 3 or 4 words correct could easily be random noise. With a 50-word list, the confidence interval tightens. Updated tables using computer simulation found that the boundaries shifted modestly from the older versions, with most critical ranges narrowing slightly.6PubMed. Critical difference table for word recognition testing derived using computer simulation The practical point is that small score fluctuations between appointments do not necessarily mean your hearing is getting better or worse. Ask your audiologist whether the change exceeds the critical difference before worrying about it.
Longer word lists improve precision, but clinics face time pressure. Presenting multiple 50-word lists to hunt for PB Max can actually inflate the score, because the listener starts recognizing patterns or gets extra practice with the test format.7PubMed. Clinical Strategies for Sampling Word Recognition Performance So there is a practical tension between wanting more data and not wanting the test itself to skew the results.
The Quiet-Room Problem
Perhaps the biggest limitation of the standard WRS is that it happens in silence. Words are delivered one at a time, in a soundproof booth, with no competing noise. Real life is nothing like that. You are trying to follow conversation at a restaurant, on a bus, or over the din of a family dinner. And the gap between quiet-booth scores and real-world performance can be enormous.
One clinical dataset illustrated this starkly: roughly 70% of listeners tested had word recognition scores of 80% or higher in quiet, yet only about 7% of the same group performed normally on a speech-in-noise test.8PubMed Central. Speech-in-Noise Testing: An Introduction for Audiologists That is a massive disconnect. A clinician looking only at the quiet score might conclude your hearing is “pretty good,” while you are genuinely struggling to understand people in everyday settings.
Research has confirmed that you cannot reliably predict speech-in-noise ability from either the audiogram or quiet word recognition scores.9PubMed. Normal and hearing-impaired word recognition scores for monosyllabic words in quiet and noise Two people with identical audiograms and identical WRS scores can perform very differently when background noise enters the picture. This has led to a growing push to supplement or even replace the quiet WRS with standardized speech-in-noise tests, such as the QuickSIN or Words-in-Noise (WIN) test. A study comparing the two approaches found that speech-in-noise measures detected problems at earlier stages of hearing loss than quiet word recognition did, and that for many patients, the speech-in-noise score provided enough information that quiet testing was unnecessary.10PubMed Central. Preliminary Guidelines for Replacing Word-Recognition in Quiet With Speech in Noise Assessment in the Routine Audiologic Test Battery
None of this means the quiet WRS is useless. It remains valuable for comparing ears, tracking changes over time, flagging neural problems, and making decisions about hearing aids and cochlear implants. But if your audiologist has never tested you with background noise, and you feel like you hear fine in the office but terribly at dinner, you are not imagining things. The quiet test may simply not be capturing your real difficulty.
Rollover and Retrocochlear Red Flags
In most people, word recognition scores climb as the volume increases, then plateau. But in some cases, scores actually drop when words are presented at higher intensities. This phenomenon is called rollover, and it has historically been associated with problems beyond the inner ear, particularly tumors on the hearing nerve known as vestibular schwannomas (sometimes called acoustic neuromas).11PubMed Central. Evaluation of Asymmetries in Speech-in Noise Abilities in Audiologic Screening for Vestibular Schwannoma A rollover ratio above 0.45 was considered a useful marker for separating inner-ear problems from nerve-related ones. In practice, though, the full rollover test requires presenting words at many different levels, making it time-consuming and uncommon in routine clinics. Audiologists today are more likely to flag unexpectedly low scores in one ear compared to the other and recommend imaging, rather than run a formal rollover protocol.
Auditory neuropathy spectrum disorder (ANSD) is another condition where the WRS becomes diagnostically revealing. People with ANSD can have near-normal pure-tone hearing but dramatically poor word recognition, because the neural timing needed to decode speech is disrupted.12PubMed Central. Auditory Neuropathy Spectrum Disorders: From Diagnosis to Treatment: Literature Review and Case Reports The mismatch between a decent audiogram and an abysmal WRS is one of the hallmarks that points clinicians toward this diagnosis.
What WRS Means for Hearing Aids and Cochlear Implants
WRS results heavily influence treatment decisions. For hearing aid fittings, a high score suggests that amplification should work well: the brain can still decode speech sounds, it just needs them louder. A low score suggests that even with a hearing aid cranking up the volume, understanding may remain limited because the issue is clarity, not just volume.
An important wrinkle here is that WRS obtained through earphones in the booth does not always predict how well you will do wearing actual hearing aids. One study found that the correlation between earphone scores and aided performance was not statistically significant, while the correlation between unaided free-field scores and aided scores was moderately strong.13PubMed Central. Clinical implications of word recognition differences in earphone and aided conditions Another study found that aided WRS measured with linear amplification at above-conversational levels was the single best predictor of real-world hearing aid outcomes, outperforming unaided earphone scores and audiogram-based predictions.14PubMed. Predicting Aided Outcome With Aided Word Recognition Scores Measured With Linear Amplification at Above-conversational Levels The message for patients: if your audiologist measures your WRS while you are actually wearing your hearing aids, that result is more informative about your daily experience than the score from the headphone test.
For cochlear implant candidacy, WRS is one of the primary gatekeepers. Traditional guidelines required scores below 40–50% in the best-aided ear before a patient could be considered. But a growing body of evidence suggests those thresholds are too strict. One study tracked adults with preoperative best-aided monosyllabic word scores averaging 41% who went on to receive cochlear implants. Postoperatively, they gained an average of 27 percentage points with the implant alone and 40 points when combining the implant with a hearing aid on the other ear.15PubMed Central. Evidence for the expansion of adult cochlear implant candidacy
Researchers have also questioned the common assumption that patients with earphone scores of 50% or higher are not cochlear implant candidates. When these patients were fitted with hearing aids and tested in aided conditions, their aided scores were on average about 36 percentage points lower than their earphone scores. More than 60% of this group scored below 50% on aided sentence recognition in quiet, and over 80% scored below 50% in noise, meaning they would have qualified for a cochlear implant evaluation if aided testing had been done.16PubMed Central. Earphone and Aided Word Recognition Differences in Cochlear Implant Candidates This is a case where a single earphone WRS number can create a bottleneck, keeping people from treatments that would help them, simply because the test conditions do not reflect how they actually function day to day. The push to revise candidacy criteria from sentence-level tests to word-level tests, and from earphone to aided conditions, continues to gain traction in the field.17PubMed Central. Evaluation of a Revised Indication for Determining Adult Cochlear Implant Candidacy
When Aging and Cognition Complicate the Picture
Hearing loss and cognitive decline are tangled together in older adults, and WRS sits right at the intersection. Understanding a spoken word is not a purely auditory task; it demands working memory, attention, and the ability to use sentence context to fill in gaps. Research on older listeners found that verbal working memory and age were among the strongest predictors of overall listening performance, even after accounting for hearing loss.18PubMed. Working memory affects older adults’ use of context in spoken-word recognition Older adults with stronger working memory were better at using contextual cues to recognize words, while those with weaker working memory struggled more, especially in noisy conditions.
This means that a low WRS in an older adult does not always point to the ear alone. Cognitive slowing, reduced working memory capacity, or early neurodegenerative changes can all drag scores down. Audiologists who work with older patients are increasingly aware that a hearing aid prescription based purely on the audiogram and WRS may not fully address the problem if cognitive contributions are not considered.
Testing Non-Native Speakers
Standard WRS word lists were designed for native English speakers. If English is your second language, the test becomes harder for reasons that have nothing to do with your hearing. Non-native listeners have been shown to struggle more with words that require fine phonetic discrimination at the sound-segment level, even when they are familiar with the vocabulary.19PubMed Central. Recognition of spoken words by native and non-native listeners: talker-, listener-, and item-related factors The challenge becomes especially pronounced in noise. Research comparing bilingual listeners tested in both their first language and English found that the effect of background noise was largest when testing in the second language; hearing-impaired listeners were not disadvantaged in their first language but scored significantly lower in English.20Canadian Journal of Speech-Language Pathology and Audiology. Speech audiometry on non-native speakers of English
Some countries have developed their own word lists in the local language, but many clinics worldwide still rely on English-language materials even when patients are not native speakers. If you are being tested in a language that is not your strongest, it is worth mentioning that to your audiologist. A lower score might reflect language familiarity rather than a hearing problem, and the distinction changes what treatment makes sense.
Testing Children
Young children cannot simply repeat back a list of words the way adults can, so pediatric versions of word recognition testing use modified formats. A common approach is a closed-set procedure, where the child sees pictures on a screen and points to the one matching the word they heard, rather than repeating it aloud. One validated method presents the child with three pictures and asks them to point; the target words vary in either their consonant or vowel content, allowing audiologists to separately assess how well a child perceives those two categories of speech sounds.21PubMed Central. Can Closed-Set Word Recognition Differentially Assess Vowel and Consonant Perception for School-Age Children With and Without Hearing Loss? This kind of granularity is useful for guiding therapy and hearing device programming in children, since a child who hears vowels well but misses consonants needs different support than one who struggles with both.
Automated Scoring and the Future of the Test
Traditionally, a human audiologist listens to your response and scores it in real time, which introduces some subjectivity, especially for borderline responses. Recent work on automated speech recognition systems has shown that deep neural networks can score spoken responses with accuracy statistically similar to expert human scorers, both in quiet and in noisy conditions, and with comparable test-retest reliability.22PubMed Central. Automating Speech Audiometry in Quiet and in Noise Using a Deep Neural Network Separate research on automated word and sentence correction algorithms found that the automated systems performed as well as, and in some cases better than, human operators, opening the door to remote testing over the internet.23PubMed. Automatic testing of speech recognition
This technology could make WRS testing more accessible and more consistent, particularly for people in rural areas or countries where audiologists are scarce. It could also standardize scoring in research studies, where inter-scorer variability has been a long-standing source of noise in the data. The fundamental test, you hear a word and try to repeat it, is unlikely to change. But the infrastructure around it is shifting toward greater automation, remote delivery, and eventually integration with speech-in-noise measures that better reflect how you actually experience hearing loss in your daily life.